All electronic components consume some energy to provide their functions. Operating efficiency is a key measure of interest, as this directly impacts operating costs. For AC to DC power supplies, efficiency is a comparison of input power consumed versus output power produced. A power supply requiring 100 watts of AC input power to produce 90 watts of DC output is considered 90% efficient. The more efficient the power supplies are, the less it costs to operate them.
The operating efficiency of the power supply, such as Switched Mode Rectifier (SMR), varies with output load. The “sweet spot” where the power supplies operate most economically usually occurs with output load levels between 50% and 70% of rated capacity. Provisioning guidelines for power equipment often utilize worst-case scenarios that can cause the power plant to be underutilized in daily operation. Operating power supplies at very low output power levels can decrease the efficiency, therefore increasing operating costs.
In addition, it is widely recognized in the world of Data Centers that servers are typically operated at less than 25% of the rated power supplies used in these servers. It is also well known that power supplies are not at its best efficiency when operating at 25% of their rated power. Today's power supplies are most efficient when operate approximately between 40% and 80% of their rated load.
When a server is designed, a power budget is created. The power budget includes the worst case power consumptions for all components from all approved component suppliers. For example, a 2 Gb memory stick from supplier A might draw 6 W. A similar 2 Gb one from supplier B might draw 8 W. That is 25% difference in power consumption. Similar conditions exist for Hard Drives, PCI cards as well as individual components on the server Mother Board.
The power budget would have to include the highest power consumption to assure sufficient power availability for the servers. There are components that are not yet available at the time of design but might be available during the life of the servers for upgrading purposes. Examples are the servers might be designed with 2 GHz with 1000 MHz Front-Side-Bus today, but it must accommodate a future upgrade to 2.2 GHz with 1333 MHz FSB a year down the road. The power budget must account for future upgrade for sufficient power availability.
These servers are general purposes servers which designed to serve a wide range of users from extreme computing extensive applications to lesser computing data bases. The servers may have been designed to accommodate 2 processor sockets of various speeds, 12 slots for memory sticks of up to 4 Gb each, 6 slots for hard drives of up to 250 Gb each and 4 slots for PCI cards for up to 25 W per card.
Only a small percentage of users are capable of taking advantages of these fully load servers. Typical servers might only use half of the available memory slots, hard drive slots or that of PCI slots. Typical servers might use 2 Gb or less memory sticks instead of the 4 Gb ones, 160 Gb or less hard drives instead of 250 Gb ones, or 12 W PCI cards instead of the 25 W cards.
A plurality of power supplies might be designed for providing power to multiple servers that have different loading capability. The power supplies might also be designed for multiple server generations. Many servers naturally are well under-utilizing their power availability. All of this adds up to the typical servers draw less than 50% of the available power from their rated power supplies.
In addition to the under-utilizing, many Data Centers also uses power supply redundancy as a mean to guard against possible power supply failures or loss of facility power. Please refer to FIG. 1. Power supply redundancy is the use of two power supplies 11, 12 in the same server 1a to provide reliability. Each power supply 11, 12 is hot-swappable in the same server 1a. Each power supply 11, 12 is capable of supplying power to the server 1a on its own. In the event of one power supply 11, 12 fails or one input voltage from facility AC1, AC2 is loss, the remaining power supply 11, 12 with its remaining input voltage from facility AC1, AC2 will keep the server 1a operating while the failed power supply 11, 12 can be replaced or the input voltage from facility AC1, AC2 can be restored. As shown in FIG. 1, the use of power supply redundancy brings the operating of the power supply 11, 12 to less than 25% from less than 50% of its rated power capability as power supply 10.
FIG. 2 shows typical power efficiency profile of power supply with respect to delivered power. At 25% load the power supply efficiency is at 85% as compared to its efficiency of 89% at 50% load.
FIGS. 3A and 3B show two similar servers operating side by side with redundancy power supplies. With servers lb, ic operate at less than 50% of the rated power supplies, and the power supplies 13, 14, 15, 16 themselves operate at less than 25% of their own rated capability. Operating power supplies at very low output power levels can decrease the efficiency, therefore increasing operating costs.
It is desired to develop a power management system capable of saving power and optimizing operating efficiency of the power supplies for providing power with back-up or redundancy to plural loads, such as servers, computer systems, telecom equipments, etc.